Ethics in plotting

Responsibility

• The first goal is honest communication
• Accurately and realistically represent the effect (or lack of)
• Match your statistics
• Digestible version often leaves out important details

Readings

• Data science ethics in Modern Data Science with R by Baumer, Kaplan, and Horton
• The Ethics of Data Visualization

Example

Two variables with a lot of samples. Similar standard deviations. Means differ by 0.5%

# A tibble: 10,000 × 2
      x1    x2
   <dbl> <dbl>
 1  90.1  90.0
 2  89.9  90.3
 3  90.1  90.2
 4  90.3  90.2
 5  89.7  90.7
 6  90.2  90.5
 7  90.3  90.6
 8  90.2  90.2
 9  90.0  90.8
10  90.3  90.4
# … with 9,990 more rows

# A tibble: 1 × 4
  x1_Mean x1_SD x2_Mean x2_SD
    <dbl> <dbl>   <dbl> <dbl>
1    90.0 0.252    90.5 0.251

Two ways to plot the same data

Always show as much of the raw data as you can

Dynamite plots must die

Always show as much of the raw data as you can

Dynamite plots must die

Always show as much of the raw data as you can

Dynamite plots must die

Always show as much of the raw data as you can

Dynamite plots must die

The complexity - communication challenge